Robust visual tracking via weighted spatio-temporal context learning

Implementing a robust visual tracker is a challenging task due to many disturbing factors such as illumination changes, appearance changes, rotation, partial or full occlusion, etc. The local context surrounding of the target could provide much effective information in getting a robust tracker. The spatio-temporal context (STC) learning algorithm proposed recently considers the information of the dense context around the target and has achieved a better performance. However, STC treats the whole region of the context equally, which weakens the effectiveness of the context information. In this paper, we propose a novel weighted spatio-temporal context (WSTC) learning algorithm. Our algorithm considers the surrounding context discriminatively and incorporates a weighted matrix by evaluating the motion consistencies of different regions with the tracking target. Extensive experimental results on public benchmark databases show that our algorithm outperforms the original STC algorithm and other state-of-the-art algorithms. Copyright © 2015 Acta Automatica Sinica. All rights reserved.